Welcome!
Nano-scale transistors fill warehouse-scale supercomputers, yet their performance still constrains development of the jets that defend us, the medical therapies our lives depend upon, and the energy sources that will power our generation into the next.
We’re the Computational Physics Group at Georgia Tech. We develop computational models and numerical methods for these applications. Our methods buttress algorithms crafted for efficient use of the latest supercomputers and their architectures. We develop open-source software for these methods that scales to the world’s largest supercomputers.
In August 2025 our group conducted the largest-ever CFD simulation at 200T grid points (1 quadrillion degrees of freedom) on OLCF Frontier and LLNL El Capitan without loss of accuracy. This work was a 2025 Gordon Bell Prize finalist.
Check out our papers to learn more.
Openings? Visit this page if you’re interested in joining our group.
Examples of our work
Multiphase flow problems at the core of biological, energy, naval, and aerodynamic problems. We developed an implementation of the IGR technique with Florian Schäfer for simulating these flows. In August 2025 we set the record for the largest CFD simulation at 1 quadrillion degrees of freedom (200T grid points) for simulating these phenomena, using the entire OLCF Frontier system. MFC, an open-source solver we maintain, demonstrates such scale-resolving simulation of a multi-rocket-booster configuration above (viz. via Ph.D. student Ben Wilfong).
The spectral boundary integral method leads to high-fidelity prediction and analysis of blood cells transitioning to chaos in a microfluidic device. This method of simulation provides resolution of strong cell membrane deformation with scant computational resources. We developed a stochastic model for the cell-scale flow, enabling microfluidic device design and improving treatment outcomes. The video above shows a microaneurysm (viz. via student Suzan Manasreh).
News
December 4, 2025 Georgia Tech runs a story on our rocket simulations, as discussed heretofore. This is closely related to the ACM 2025 Gordon Bell Finalist work with former GT faculty, new at the Courant Institute, Prof. Florian Schaefer.
November 26, 2025 Our group collaborated with Diego Vaca-Revelo and Aswin Gnanaskandan of WPI on the hardware acceleration of bubbly flow models in Navier-Stokes solvers, demonstrated in MFC. The preprint is available here.
November 24, 2025 Our group, along with collaborators at UT-Austin, Michigan, and Brown, have pushed a preprint on Bayesian model selection for soft materials at very high strain rates. It’s available here!
November 22, 2025 We are at APS DFD 2025 in Houston for the next few days. Ben, Spencer, and Tianyi are giving talks. Say hello if you see us!
November 19, 2025 Lawrence Livermore National Lab ran a nice story about our work on the largest and fastest CFD simulation (in particular on El Capitan), which is a 2025 ACM Gordon Bell Finalist. NVIDIA also posted about us in a blog post on their website.
November 15, 2025 Much of the group is at SC25 this week. Spencer is giving invited talks at the OpenACC BoF and the Gordon Bell Prize presentations on Tuesday and the NVIDIA Booth on Wednesday. Ben Wilfong is giving a paper talk at the HPCTESTS workshop.
October 25, 2025 Spencer published a paper in Computers and Fluids. It presents a scheme that can integrate for the moments of radial bubble dynamics with time-independent spectral accuracy.
October 21, 2025 Group Ph.D. student Haocheng Yu leads work that now appears on arXiv. Collaboration with Profs. L. Sankar and K. Ahuja in AE. The work focuses on high amplitude sound through good but still imperfect ear plugs, and the physical mechanisms that maintain them.
October 20, 2025 NCSA has written a piece on the group and our collaborator’s work towards the Gordon Bell Prize and its use of the Grace Hopper architecture. Find it here.
October 16, 2025 Spencer presents our work, Symbolic Computational Representations of Combustion Thermochemistry, at the 2025 International Conference on Numerical Combustion in Rome, Italy. Part of a collaboration with Prof. Esteban Cisneros-Garibay at the University of Tennessee Space Institute; the work is also conducted by CSE Ph.D. student Dimitrios Adam and (nominally) Spencer. We achieve near-peak compute performance on diverse accelerators for diverse flavors of chemical reactions via optimal symbolic traversal of the computational graph.
